An Industrial i-Butane/n-Butane Fractionator

Klemola and Ilme [Ind. Eng. Chem., 35, 4579 (1996)] and Ilme [Ph.D. Thesis, University of
Lapeenranta, Finland (1997)] report data from an industrial i-butane/n-butane fractionator that is
used here as the basis for this case study.

The column has 74 valve trays, the feed was introduced onto tray 37. The key design parameters
for the valve trays are given in the table below.

Column Height 51.8m Downcomer Area (center) 0.86 m2

Column Diameter 2.9m Tray Spacing 0.6m
Number of Trays 74 Hole Diameter 39mm
Weir Length (side) 1.859m Total Hole Area 0.922 m2
Weir Length (center) 2.885m Outlet Weir Height 51mm
Liquid Flowpath Length 0.967m per pass Tray Thickness 2mm
2
Active Area 4.9 m Number of Valves per Tray 772
2
Downcomer Area (side) 0.86 m Free Fractional Hole Area 18.82%

The measured compositions and flow rates of the feed and products for the C4 splitter are
summarized in the table below.

Measured Feed and Product Flows and Compositions (mass %)

for i-Butane/n-Butane Fractionator (Ilme, 1997)

Species Feed Top Bottom

Propane 1.50 5.30 0.00
Isobutane 29.4 93.5 0.30
n-Butane 67.7 0.20 98.1
C4 olefins 0.50 1.00 0.20
Neopentane 0.10 0.00 0.20
Isopentane 0.80 0.00 1.10
n-Pentane 0.10 0.00 0.10
Total flow (kg/h) 26234 8011 17887

Other measured parameters are as follows:

Other details of the i-Butane/n-Butane Fractionator

Reflux Flow Rate kg/h 92838

Reflux Temperature C 18.5
Column Top Pressure kPa 658.6
Pressure drop per tray kPa 0.47
Feed Pressure kPa 892.67
Boiler Duty MW 10.24
Rarely, and this is a case in point, are plant data in exact material balance and it will be necessary
to reconcile errors in such measurements before continuing. The feed and product compositions
as adjusted by Ilme so that they satisfy material balance constraints, are provided below. Note
how the C4 olefins are assigned to isobutene and 1-butene.

Adjusted feed and product compositions (mass %) and flows

for i-Butane/n-Butane Fractionator (Ilme, 1997)

Species Feed Top Bottom

Propane 1.54 4.94 0.00
Isobutane 29.5 94.2 0.3
n-Butane 67.7 0.20 98.1
Isobutene 0.13 0.23 0.08
1-butene 0.20 0.41 0.10
Neopentane 0.11 0.00 0.17
Isopentane 0.77 0.00 1.12
n-Pentane 0.08 0.00 0.11
Total flow (kg/h) 26122 8123 17999

To proceed with building a model of this column we specify the number of stages equal to the
number of trays plus condenser and reboiler (N = 76). The common arrangement of locating the
actual feed between stages may need modeling as two separate feeds; the liquid portion to the
stage below and the vapor portion to the stage above. In this particular illustration the feed is
(assumed to be) saturated liquid and we provide just a single feed to stage 38.

Upon computing the bubble point of the overhead product we will find that the measured reflux
temperature is well below the estimated boiling point. Thus, we choose the subcooled condenser
model. The steady-state concept of the so-called sub-cooled condenser often does not exist in
practice. Instead, the condenser is in vapor-liquid equilibrium with the vapor augmented by a
blanket of non-condensable gas (that has the effect of lowering the dew point of the overhead
vapor). The sub-cooled condenser is a convenient work around for steady-state models (as is
needed here), but not for dynamic models. We assume a partial reboiler.

The specifications made to model this column are summarized below:

Variable Number Value

Number of stages 1 N = 76
Feed stage location 1 39
Component flows in feed c=8 See other table
Feed pressure 1 120 psia
Feed vapor fraction 1 0
Pressure at the top of the column 1 658.6 kPa
Pressure drop per stage N 1 = 75 0.47 kPa
Heat duty on each stage except reboilers and condensers N - 2 = 74 Qj = 0
Reflux ratio (replaces heat duty of condenser) 1 R = 11.588
 Bottoms flow rate (replaces heat duty of reboiler) 1 B = 17999 kg/h
Temperature of reflux 1 291.65 K
Total 165

Finally, we must select appropriate methods of estimating thermodynamic properties. Ilme

(1997) used the SRK equation of state to model this column, whereas Klemola and Ilme (1996)
had earlier used the UNIFAC model for liquid phase activity coefficients, the Antoine equation
for vapor pressures and the SRK equation for vapor phase fugacities only. For this exercise we
used the Peng-Robinson equation of state. Computed product compositions and flow rates
are shown in the table below.

Specified feed (Ilme, 1997) and computed product compositions (mass %) and flows
for i-Butane/n-Butane Fractionator

Compound Feed Top Bottom

Propane 1.54 4.95 0.00
Isobutane 29.49 93.67 0.53
n-Butane 67.68 0.73 97.89
Isobutene 0.13 0.29 0.06
1-butene 0.20 0.36 0.13
Neopentane 0.11 0.00 0.16
Isopentane 0.77 0.00 1.12
n-Pentane 0.08 0.00 0.12
Total flow (kg/h) 26122 8123.01 17999

The agreement with the adjusted material balance (tabulated above) appears to be quite good and
to a first approximation it seems that we have a good model of the column.

It must be noted that although this column is distilling a mixture containing at least 8 identifiable
compounds, only two are present in significant amounts and, therefore, this is essentially a
binary separation. It is usually relatively straightforward to match product compositions in
processes involving only two different species simply by adjusting the number of equilibrium
stages. We shall return to this point later.

It is possible to estimate the overall efficiency for a column such as this one simply by adjusting
the number of equilibrium stages in each section of the column that are needed to match the mass
fractions of i-butane in the distillate and n-butane in the bottoms. Using the SRK equation of
state for estimating thermodynamic properties Ilme (1997) found that 82 equilibrium stages (plus
condenser and reboiler) and the feed to stage 38 were required. This corresponds to an overall
column efficiency of 82/74 = 111%. Klemola & Ilme (1996) used the UNIFAC model for liquid
phase activity coefficients, the Antoine equation for vapor pressures and the SRK equation for
vapor phase fugacities only and found that 88 ideal stages were needed; this corresponding to an
overall efficiency of 119%. With the Peng-Robinson equation of state for the estimation of
thermodynamic properties we find that 84 stages are needed (while maintaining the feed to the
center stage as is the case here); the overall column efficiency for this model being 114%. The
differences between these efficiencies are not large in this case, but the important point here is
that efficiencies all types depend on the choice of model used to estimate thermodynamic
properties. Caution must, therefore, be exercised when using efficiencies determined in this way
to predict column performance.

As an alternative to varying the number of stages we may prefer to maintain a one-to-one

correspondence between the number of stages and the number of actual trays, 74 in this case
(plus condenser and reboiler), with the feed to tray 38. Using the Peng-Robinson equation of
state and a Murphree stage efficiency of 116% we find the product mass fractions that are in
excellent agreement with the plant data. The McCabe-Thiele diagram for this case, assembled
from the results of the simulation, is shown below

McCabe-Thiele diagram for Isobutane - n-Butane

1
Y Isobutane/(Isobutane+n-Butane)

0.8

0.6

0.4
39

0.2

0 ChemSep
0 0.2 0.4 0.6 0.8 1
X Isobutane/(Isobutane+n-Butane)
McCabe-Thiele diagram for C4 splitter
Composition profiles computed from this model are shown below. Note that the mole fractions
are shown on a logarithmic axis so that all of the composition profiles can easily be seen.

20
Stage

40

60

1e-05 0.0001 0.001 0.01 0.1 1

Liquid phase mole fraction

X Propane X Isobutene X Isopentane

X Isobutane X 1-butene X n-Pentane
X n-Butane X Neopentane
Liquid phase mole fraction profiles for
i-Butane/n-Butane Fractionator

It must be remembered that this is essentially a binary separation and that it is usually relatively
straightforward to match product compositions in processes involving only two different species.
In other cases involving a greater number of species with significant concentrations it will likely
be necessary to vary both the number of stages and the component efficiencies to match plant
data. We do not recommend adjusting thermodynamic model parameters in order to fit plant data
since this can have unfortunate consequences on the prediction of product distributions, process
temperatures and/or pressures.

When we create a nonequilibrium model of this or any column we do not need to guess how
many stages to use in each section of the column. The real column had 74 valve trays; the model
column includes 74 model trays with the feed to tray 38 (plus a (subcooled) condenser and a
reboiler, both of which are modeled as equilibrium stages as described above). All operating
specifications are the same as for the corresponding equilibrium stage model. It is necessary to
choose models that allow for the estimation of the rates of interphase mass transfer; that means
selecting vapor and liquid flow models and correlations to estimate the mass transfer coefficients
in each phase as discussed above. In this case the AIChE correlations were used. It is known that
this method is more conservative than others (i.e. the predicted efficiencies are lower). The
importance of the flow model is clear from the simulation results tabulated below. The predicted
component Murphree efficiencies vary more widely from stage to stage and from component to
component than might be expected for a system like this. The Baur efficiency, on the other hand,
does not change by more than a few percentage points over the height of the column; the value in
the table below is an average of that computed for each tray from the simulation.

Vapor flow model Liquid flow model iC4 in Distillate (%) nC4 in Bottoms(%) Efficiency(%)
Mixed Mixed 90.2 96.3 63
Plug Mixed 92.2 97.2 78
Plug Dispersion 93.9 98.0 106

Internal vapor and/or liquid composition data rarely is available, but such data is the best
possible for model discrimination and validation. It is often relatively easy to match even a
simple model only to product compositions. In the absence of composition profiles, the internal
temperature profile can often be as useful provided that it is known to which phase a measured
temperature pertains. The table below compares the few available measured tray temperatures
with those computed during the simulation. The agreement is quite good.

Tray Temperature (oC)

Measured Predicted
9 47.5 48.6
65 62.2 62.5
74 63.2 63.1

A portion of the McCabe-Thiele diagram for the simulation involving plug flow of vapor and
dispersion flow of the liquid is shown below. For a nonequilibrium column these diagrams can
only be constructed from the results of a computer simulation. Note that the triangles that
represent the stages extend beyond the curve that represents the equilibrium line; this is because
the efficiencies are greater than 100%.
0.95
Y Isobutane/(Isobutane+n-Butane)

0.9
0.9 0.95
X Isobutane/(Isobutane+n-Butane)
Expanded view of upper right corner of McCabe-Thiele diagram for C4 splitter.
In this particular case the converged composition and temperature profiles have the same shape
as those obtained with the equilibrium stage model (with specified efficiency) and, therefore, are
not shown. The reason for the similarity is that, as noted above, this is basically a binary
separation of very similar compounds. The important point here is that, unlike the equilibrium
stage model simulations, the nonequilibrium model predicted how the column would perform; no
parameters were adjusted to provide a better fit to the plant data. That is not to say, of course,
that NEQ models cannot be used to fit plant data. In principle, the mass transfer coefficients and
interfacial area (or parameters in the equations used to estimate them) can be tuned to help the
model better fit plant data.